2 * NET3 IP device support routines.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
9 * Derived from the IP parts of dev.c 1.0.19
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk>
15 * Alan Cox, <gw4pts@gw4pts.ampr.org>
16 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
19 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
21 * Cyrus Durgin: updated for kmod
22 * Matthias Andree: in devinet_ioctl, compare label and
23 * address (4.4BSD alias style support),
24 * fall back to comparing just the label
29 #include <asm/uaccess.h>
30 #include <linux/bitops.h>
31 #include <linux/capability.h>
32 #include <linux/module.h>
33 #include <linux/types.h>
34 #include <linux/kernel.h>
35 #include <linux/string.h>
37 #include <linux/socket.h>
38 #include <linux/sockios.h>
40 #include <linux/errno.h>
41 #include <linux/interrupt.h>
42 #include <linux/if_addr.h>
43 #include <linux/if_ether.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <linux/etherdevice.h>
47 #include <linux/skbuff.h>
48 #include <linux/init.h>
49 #include <linux/notifier.h>
50 #include <linux/inetdevice.h>
51 #include <linux/igmp.h>
52 #include <linux/slab.h>
53 #include <linux/hash.h>
55 #include <linux/sysctl.h>
57 #include <linux/kmod.h>
58 #include <linux/netconf.h>
62 #include <net/route.h>
63 #include <net/ip_fib.h>
64 #include <net/rtnetlink.h>
65 #include <net/net_namespace.h>
66 #include <net/addrconf.h>
68 #include "fib_lookup.h"
70 static struct ipv4_devconf ipv4_devconf = {
72 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
73 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
74 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
75 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
76 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
77 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
81 static struct ipv4_devconf ipv4_devconf_dflt = {
83 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
86 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
87 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
88 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
89 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
93 #define IPV4_DEVCONF_DFLT(net, attr) \
94 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
96 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
97 [IFA_LOCAL] = { .type = NLA_U32 },
98 [IFA_ADDRESS] = { .type = NLA_U32 },
99 [IFA_BROADCAST] = { .type = NLA_U32 },
100 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
101 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
102 [IFA_FLAGS] = { .type = NLA_U32 },
105 #define IN4_ADDR_HSIZE_SHIFT 8
106 #define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
108 static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
109 static DEFINE_SPINLOCK(inet_addr_hash_lock);
111 static u32 inet_addr_hash(struct net *net, __be32 addr)
113 u32 val = (__force u32) addr ^ net_hash_mix(net);
115 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
118 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
120 u32 hash = inet_addr_hash(net, ifa->ifa_local);
122 spin_lock(&inet_addr_hash_lock);
123 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
124 spin_unlock(&inet_addr_hash_lock);
127 static void inet_hash_remove(struct in_ifaddr *ifa)
129 spin_lock(&inet_addr_hash_lock);
130 hlist_del_init_rcu(&ifa->hash);
131 spin_unlock(&inet_addr_hash_lock);
135 * __ip_dev_find - find the first device with a given source address.
136 * @net: the net namespace
137 * @addr: the source address
138 * @devref: if true, take a reference on the found device
140 * If a caller uses devref=false, it should be protected by RCU, or RTNL
142 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
144 u32 hash = inet_addr_hash(net, addr);
145 struct net_device *result = NULL;
146 struct in_ifaddr *ifa;
149 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash) {
150 if (ifa->ifa_local == addr) {
151 struct net_device *dev = ifa->ifa_dev->dev;
153 if (!net_eq(dev_net(dev), net))
160 struct flowi4 fl4 = { .daddr = addr };
161 struct fib_result res = { 0 };
162 struct fib_table *local;
164 /* Fallback to FIB local table so that communication
165 * over loopback subnets work.
167 local = fib_get_table(net, RT_TABLE_LOCAL);
169 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
170 res.type == RTN_LOCAL)
171 result = FIB_RES_DEV(res);
173 if (result && devref)
178 EXPORT_SYMBOL(__ip_dev_find);
180 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
182 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
183 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
186 static void devinet_sysctl_register(struct in_device *idev);
187 static void devinet_sysctl_unregister(struct in_device *idev);
189 static void devinet_sysctl_register(struct in_device *idev)
192 static void devinet_sysctl_unregister(struct in_device *idev)
197 /* Locks all the inet devices. */
199 static struct in_ifaddr *inet_alloc_ifa(void)
201 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
204 static void inet_rcu_free_ifa(struct rcu_head *head)
206 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
208 in_dev_put(ifa->ifa_dev);
212 static void inet_free_ifa(struct in_ifaddr *ifa)
214 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
217 void in_dev_finish_destroy(struct in_device *idev)
219 struct net_device *dev = idev->dev;
221 WARN_ON(idev->ifa_list);
222 WARN_ON(idev->mc_list);
223 kfree(rcu_dereference_protected(idev->mc_hash, 1));
224 #ifdef NET_REFCNT_DEBUG
225 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
229 pr_err("Freeing alive in_device %p\n", idev);
233 EXPORT_SYMBOL(in_dev_finish_destroy);
235 static struct in_device *inetdev_init(struct net_device *dev)
237 struct in_device *in_dev;
241 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
244 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
245 sizeof(in_dev->cnf));
246 in_dev->cnf.sysctl = NULL;
248 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
249 if (!in_dev->arp_parms)
251 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
252 dev_disable_lro(dev);
253 /* Reference in_dev->dev */
255 /* Account for reference dev->ip_ptr (below) */
258 devinet_sysctl_register(in_dev);
259 ip_mc_init_dev(in_dev);
260 if (dev->flags & IFF_UP)
263 /* we can receive as soon as ip_ptr is set -- do this last */
264 rcu_assign_pointer(dev->ip_ptr, in_dev);
273 static void in_dev_rcu_put(struct rcu_head *head)
275 struct in_device *idev = container_of(head, struct in_device, rcu_head);
279 static void inetdev_destroy(struct in_device *in_dev)
281 struct in_ifaddr *ifa;
282 struct net_device *dev;
290 ip_mc_destroy_dev(in_dev);
292 while ((ifa = in_dev->ifa_list) != NULL) {
293 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
297 RCU_INIT_POINTER(dev->ip_ptr, NULL);
299 devinet_sysctl_unregister(in_dev);
300 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
303 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
306 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
309 for_primary_ifa(in_dev) {
310 if (inet_ifa_match(a, ifa)) {
311 if (!b || inet_ifa_match(b, ifa)) {
316 } endfor_ifa(in_dev);
321 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
322 int destroy, struct nlmsghdr *nlh, u32 portid)
324 struct in_ifaddr *promote = NULL;
325 struct in_ifaddr *ifa, *ifa1 = *ifap;
326 struct in_ifaddr *last_prim = in_dev->ifa_list;
327 struct in_ifaddr *prev_prom = NULL;
328 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
332 /* 1. Deleting primary ifaddr forces deletion all secondaries
333 * unless alias promotion is set
336 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
337 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
339 while ((ifa = *ifap1) != NULL) {
340 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
341 ifa1->ifa_scope <= ifa->ifa_scope)
344 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
345 ifa1->ifa_mask != ifa->ifa_mask ||
346 !inet_ifa_match(ifa1->ifa_address, ifa)) {
347 ifap1 = &ifa->ifa_next;
353 inet_hash_remove(ifa);
354 *ifap1 = ifa->ifa_next;
356 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
357 blocking_notifier_call_chain(&inetaddr_chain,
367 /* On promotion all secondaries from subnet are changing
368 * the primary IP, we must remove all their routes silently
369 * and later to add them back with new prefsrc. Do this
370 * while all addresses are on the device list.
372 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
373 if (ifa1->ifa_mask == ifa->ifa_mask &&
374 inet_ifa_match(ifa1->ifa_address, ifa))
375 fib_del_ifaddr(ifa, ifa1);
380 *ifap = ifa1->ifa_next;
381 inet_hash_remove(ifa1);
383 /* 3. Announce address deletion */
385 /* Send message first, then call notifier.
386 At first sight, FIB update triggered by notifier
387 will refer to already deleted ifaddr, that could confuse
388 netlink listeners. It is not true: look, gated sees
389 that route deleted and if it still thinks that ifaddr
390 is valid, it will try to restore deleted routes... Grr.
391 So that, this order is correct.
393 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
394 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
397 struct in_ifaddr *next_sec = promote->ifa_next;
400 prev_prom->ifa_next = promote->ifa_next;
401 promote->ifa_next = last_prim->ifa_next;
402 last_prim->ifa_next = promote;
405 promote->ifa_flags &= ~IFA_F_SECONDARY;
406 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
407 blocking_notifier_call_chain(&inetaddr_chain,
409 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
410 if (ifa1->ifa_mask != ifa->ifa_mask ||
411 !inet_ifa_match(ifa1->ifa_address, ifa))
421 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
424 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
427 static void check_lifetime(struct work_struct *work);
429 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
431 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
434 struct in_device *in_dev = ifa->ifa_dev;
435 struct in_ifaddr *ifa1, **ifap, **last_primary;
439 if (!ifa->ifa_local) {
444 ifa->ifa_flags &= ~IFA_F_SECONDARY;
445 last_primary = &in_dev->ifa_list;
447 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
448 ifap = &ifa1->ifa_next) {
449 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
450 ifa->ifa_scope <= ifa1->ifa_scope)
451 last_primary = &ifa1->ifa_next;
452 if (ifa1->ifa_mask == ifa->ifa_mask &&
453 inet_ifa_match(ifa1->ifa_address, ifa)) {
454 if (ifa1->ifa_local == ifa->ifa_local) {
458 if (ifa1->ifa_scope != ifa->ifa_scope) {
462 ifa->ifa_flags |= IFA_F_SECONDARY;
466 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
467 net_srandom(ifa->ifa_local);
471 ifa->ifa_next = *ifap;
474 inet_hash_insert(dev_net(in_dev->dev), ifa);
476 cancel_delayed_work(&check_lifetime_work);
477 schedule_delayed_work(&check_lifetime_work, 0);
479 /* Send message first, then call notifier.
480 Notifier will trigger FIB update, so that
481 listeners of netlink will know about new ifaddr */
482 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
483 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
488 static int inet_insert_ifa(struct in_ifaddr *ifa)
490 return __inet_insert_ifa(ifa, NULL, 0);
493 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
495 struct in_device *in_dev = __in_dev_get_rtnl(dev);
503 ipv4_devconf_setall(in_dev);
504 neigh_parms_data_state_setall(in_dev->arp_parms);
505 if (ifa->ifa_dev != in_dev) {
506 WARN_ON(ifa->ifa_dev);
508 ifa->ifa_dev = in_dev;
510 if (ipv4_is_loopback(ifa->ifa_local))
511 ifa->ifa_scope = RT_SCOPE_HOST;
512 return inet_insert_ifa(ifa);
515 /* Caller must hold RCU or RTNL :
516 * We dont take a reference on found in_device
518 struct in_device *inetdev_by_index(struct net *net, int ifindex)
520 struct net_device *dev;
521 struct in_device *in_dev = NULL;
524 dev = dev_get_by_index_rcu(net, ifindex);
526 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
530 EXPORT_SYMBOL(inetdev_by_index);
532 /* Called only from RTNL semaphored context. No locks. */
534 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
539 for_primary_ifa(in_dev) {
540 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
542 } endfor_ifa(in_dev);
546 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
548 struct net *net = sock_net(skb->sk);
549 struct nlattr *tb[IFA_MAX+1];
550 struct in_device *in_dev;
551 struct ifaddrmsg *ifm;
552 struct in_ifaddr *ifa, **ifap;
557 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
561 ifm = nlmsg_data(nlh);
562 in_dev = inetdev_by_index(net, ifm->ifa_index);
563 if (in_dev == NULL) {
568 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
569 ifap = &ifa->ifa_next) {
571 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
574 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
577 if (tb[IFA_ADDRESS] &&
578 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
579 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
582 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
586 err = -EADDRNOTAVAIL;
591 #define INFINITY_LIFE_TIME 0xFFFFFFFF
593 static void check_lifetime(struct work_struct *work)
595 unsigned long now, next, next_sec, next_sched;
596 struct in_ifaddr *ifa;
597 struct hlist_node *n;
601 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
603 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
604 bool change_needed = false;
607 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
610 if (ifa->ifa_flags & IFA_F_PERMANENT)
613 /* We try to batch several events at once. */
614 age = (now - ifa->ifa_tstamp +
615 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
617 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
618 age >= ifa->ifa_valid_lft) {
619 change_needed = true;
620 } else if (ifa->ifa_preferred_lft ==
621 INFINITY_LIFE_TIME) {
623 } else if (age >= ifa->ifa_preferred_lft) {
624 if (time_before(ifa->ifa_tstamp +
625 ifa->ifa_valid_lft * HZ, next))
626 next = ifa->ifa_tstamp +
627 ifa->ifa_valid_lft * HZ;
629 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
630 change_needed = true;
631 } else if (time_before(ifa->ifa_tstamp +
632 ifa->ifa_preferred_lft * HZ,
634 next = ifa->ifa_tstamp +
635 ifa->ifa_preferred_lft * HZ;
642 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
645 if (ifa->ifa_flags & IFA_F_PERMANENT)
648 /* We try to batch several events at once. */
649 age = (now - ifa->ifa_tstamp +
650 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
652 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
653 age >= ifa->ifa_valid_lft) {
654 struct in_ifaddr **ifap;
656 for (ifap = &ifa->ifa_dev->ifa_list;
657 *ifap != NULL; ifap = &(*ifap)->ifa_next) {
659 inet_del_ifa(ifa->ifa_dev,
664 } else if (ifa->ifa_preferred_lft !=
665 INFINITY_LIFE_TIME &&
666 age >= ifa->ifa_preferred_lft &&
667 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
668 ifa->ifa_flags |= IFA_F_DEPRECATED;
669 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
675 next_sec = round_jiffies_up(next);
678 /* If rounded timeout is accurate enough, accept it. */
679 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
680 next_sched = next_sec;
683 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
684 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
685 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
687 schedule_delayed_work(&check_lifetime_work, next_sched - now);
690 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
693 unsigned long timeout;
695 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
697 timeout = addrconf_timeout_fixup(valid_lft, HZ);
698 if (addrconf_finite_timeout(timeout))
699 ifa->ifa_valid_lft = timeout;
701 ifa->ifa_flags |= IFA_F_PERMANENT;
703 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
704 if (addrconf_finite_timeout(timeout)) {
706 ifa->ifa_flags |= IFA_F_DEPRECATED;
707 ifa->ifa_preferred_lft = timeout;
709 ifa->ifa_tstamp = jiffies;
710 if (!ifa->ifa_cstamp)
711 ifa->ifa_cstamp = ifa->ifa_tstamp;
714 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
715 __u32 *pvalid_lft, __u32 *pprefered_lft)
717 struct nlattr *tb[IFA_MAX+1];
718 struct in_ifaddr *ifa;
719 struct ifaddrmsg *ifm;
720 struct net_device *dev;
721 struct in_device *in_dev;
724 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
728 ifm = nlmsg_data(nlh);
730 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
733 dev = __dev_get_by_index(net, ifm->ifa_index);
738 in_dev = __in_dev_get_rtnl(dev);
743 ifa = inet_alloc_ifa();
746 * A potential indev allocation can be left alive, it stays
747 * assigned to its device and is destroy with it.
751 ipv4_devconf_setall(in_dev);
752 neigh_parms_data_state_setall(in_dev->arp_parms);
755 if (tb[IFA_ADDRESS] == NULL)
756 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
758 INIT_HLIST_NODE(&ifa->hash);
759 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
760 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
761 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
763 ifa->ifa_scope = ifm->ifa_scope;
764 ifa->ifa_dev = in_dev;
766 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
767 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
769 if (tb[IFA_BROADCAST])
770 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
773 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
775 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
777 if (tb[IFA_CACHEINFO]) {
778 struct ifa_cacheinfo *ci;
780 ci = nla_data(tb[IFA_CACHEINFO]);
781 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
785 *pvalid_lft = ci->ifa_valid;
786 *pprefered_lft = ci->ifa_prefered;
797 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
799 struct in_device *in_dev = ifa->ifa_dev;
800 struct in_ifaddr *ifa1, **ifap;
805 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
806 ifap = &ifa1->ifa_next) {
807 if (ifa1->ifa_mask == ifa->ifa_mask &&
808 inet_ifa_match(ifa1->ifa_address, ifa) &&
809 ifa1->ifa_local == ifa->ifa_local)
815 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
817 struct net *net = sock_net(skb->sk);
818 struct in_ifaddr *ifa;
819 struct in_ifaddr *ifa_existing;
820 __u32 valid_lft = INFINITY_LIFE_TIME;
821 __u32 prefered_lft = INFINITY_LIFE_TIME;
825 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
829 ifa_existing = find_matching_ifa(ifa);
831 /* It would be best to check for !NLM_F_CREATE here but
832 * userspace alreay relies on not having to provide this.
834 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
835 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
839 if (nlh->nlmsg_flags & NLM_F_EXCL ||
840 !(nlh->nlmsg_flags & NLM_F_REPLACE))
843 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
844 cancel_delayed_work(&check_lifetime_work);
845 schedule_delayed_work(&check_lifetime_work, 0);
846 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
847 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
853 * Determine a default network mask, based on the IP address.
856 static int inet_abc_len(__be32 addr)
858 int rc = -1; /* Something else, probably a multicast. */
860 if (ipv4_is_zeronet(addr))
863 __u32 haddr = ntohl(addr);
865 if (IN_CLASSA(haddr))
867 else if (IN_CLASSB(haddr))
869 else if (IN_CLASSC(haddr))
877 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
880 struct sockaddr_in sin_orig;
881 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
882 struct in_device *in_dev;
883 struct in_ifaddr **ifap = NULL;
884 struct in_ifaddr *ifa = NULL;
885 struct net_device *dev;
888 int tryaddrmatch = 0;
891 * Fetch the caller's info block into kernel space
894 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
896 ifr.ifr_name[IFNAMSIZ - 1] = 0;
898 /* save original address for comparison */
899 memcpy(&sin_orig, sin, sizeof(*sin));
901 colon = strchr(ifr.ifr_name, ':');
905 dev_load(net, ifr.ifr_name);
908 case SIOCGIFADDR: /* Get interface address */
909 case SIOCGIFBRDADDR: /* Get the broadcast address */
910 case SIOCGIFDSTADDR: /* Get the destination address */
911 case SIOCGIFNETMASK: /* Get the netmask for the interface */
912 /* Note that these ioctls will not sleep,
913 so that we do not impose a lock.
914 One day we will be forced to put shlock here (I mean SMP)
916 tryaddrmatch = (sin_orig.sin_family == AF_INET);
917 memset(sin, 0, sizeof(*sin));
918 sin->sin_family = AF_INET;
923 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
926 case SIOCSIFADDR: /* Set interface address (and family) */
927 case SIOCSIFBRDADDR: /* Set the broadcast address */
928 case SIOCSIFDSTADDR: /* Set the destination address */
929 case SIOCSIFNETMASK: /* Set the netmask for the interface */
931 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
934 if (sin->sin_family != AF_INET)
945 dev = __dev_get_by_name(net, ifr.ifr_name);
952 in_dev = __in_dev_get_rtnl(dev);
955 /* Matthias Andree */
956 /* compare label and address (4.4BSD style) */
957 /* note: we only do this for a limited set of ioctls
958 and only if the original address family was AF_INET.
959 This is checked above. */
960 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
961 ifap = &ifa->ifa_next) {
962 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
963 sin_orig.sin_addr.s_addr ==
969 /* we didn't get a match, maybe the application is
970 4.3BSD-style and passed in junk so we fall back to
971 comparing just the label */
973 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
974 ifap = &ifa->ifa_next)
975 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
980 ret = -EADDRNOTAVAIL;
981 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
985 case SIOCGIFADDR: /* Get interface address */
986 sin->sin_addr.s_addr = ifa->ifa_local;
989 case SIOCGIFBRDADDR: /* Get the broadcast address */
990 sin->sin_addr.s_addr = ifa->ifa_broadcast;
993 case SIOCGIFDSTADDR: /* Get the destination address */
994 sin->sin_addr.s_addr = ifa->ifa_address;
997 case SIOCGIFNETMASK: /* Get the netmask for the interface */
998 sin->sin_addr.s_addr = ifa->ifa_mask;
1003 ret = -EADDRNOTAVAIL;
1007 if (!(ifr.ifr_flags & IFF_UP))
1008 inet_del_ifa(in_dev, ifap, 1);
1011 ret = dev_change_flags(dev, ifr.ifr_flags);
1014 case SIOCSIFADDR: /* Set interface address (and family) */
1016 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1021 ifa = inet_alloc_ifa();
1024 INIT_HLIST_NODE(&ifa->hash);
1026 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
1028 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1031 if (ifa->ifa_local == sin->sin_addr.s_addr)
1033 inet_del_ifa(in_dev, ifap, 0);
1034 ifa->ifa_broadcast = 0;
1038 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1040 if (!(dev->flags & IFF_POINTOPOINT)) {
1041 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1042 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1043 if ((dev->flags & IFF_BROADCAST) &&
1044 ifa->ifa_prefixlen < 31)
1045 ifa->ifa_broadcast = ifa->ifa_address |
1048 ifa->ifa_prefixlen = 32;
1049 ifa->ifa_mask = inet_make_mask(32);
1051 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1052 ret = inet_set_ifa(dev, ifa);
1055 case SIOCSIFBRDADDR: /* Set the broadcast address */
1057 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1058 inet_del_ifa(in_dev, ifap, 0);
1059 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1060 inet_insert_ifa(ifa);
1064 case SIOCSIFDSTADDR: /* Set the destination address */
1066 if (ifa->ifa_address == sin->sin_addr.s_addr)
1069 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1072 inet_del_ifa(in_dev, ifap, 0);
1073 ifa->ifa_address = sin->sin_addr.s_addr;
1074 inet_insert_ifa(ifa);
1077 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1080 * The mask we set must be legal.
1083 if (bad_mask(sin->sin_addr.s_addr, 0))
1086 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1087 __be32 old_mask = ifa->ifa_mask;
1088 inet_del_ifa(in_dev, ifap, 0);
1089 ifa->ifa_mask = sin->sin_addr.s_addr;
1090 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1092 /* See if current broadcast address matches
1093 * with current netmask, then recalculate
1094 * the broadcast address. Otherwise it's a
1095 * funny address, so don't touch it since
1096 * the user seems to know what (s)he's doing...
1098 if ((dev->flags & IFF_BROADCAST) &&
1099 (ifa->ifa_prefixlen < 31) &&
1100 (ifa->ifa_broadcast ==
1101 (ifa->ifa_local|~old_mask))) {
1102 ifa->ifa_broadcast = (ifa->ifa_local |
1103 ~sin->sin_addr.s_addr);
1105 inet_insert_ifa(ifa);
1115 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1119 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1121 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1122 struct in_ifaddr *ifa;
1129 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1131 done += sizeof(ifr);
1134 if (len < (int) sizeof(ifr))
1136 memset(&ifr, 0, sizeof(struct ifreq));
1137 strcpy(ifr.ifr_name, ifa->ifa_label);
1139 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1140 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1143 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1147 buf += sizeof(struct ifreq);
1148 len -= sizeof(struct ifreq);
1149 done += sizeof(struct ifreq);
1155 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1158 struct in_device *in_dev;
1159 struct net *net = dev_net(dev);
1162 in_dev = __in_dev_get_rcu(dev);
1166 for_primary_ifa(in_dev) {
1167 if (ifa->ifa_scope > scope)
1169 if (!dst || inet_ifa_match(dst, ifa)) {
1170 addr = ifa->ifa_local;
1174 addr = ifa->ifa_local;
1175 } endfor_ifa(in_dev);
1181 /* Not loopback addresses on loopback should be preferred
1182 in this case. It is importnat that lo is the first interface
1185 for_each_netdev_rcu(net, dev) {
1186 in_dev = __in_dev_get_rcu(dev);
1190 for_primary_ifa(in_dev) {
1191 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1192 ifa->ifa_scope <= scope) {
1193 addr = ifa->ifa_local;
1196 } endfor_ifa(in_dev);
1202 EXPORT_SYMBOL(inet_select_addr);
1204 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1205 __be32 local, int scope)
1212 (local == ifa->ifa_local || !local) &&
1213 ifa->ifa_scope <= scope) {
1214 addr = ifa->ifa_local;
1219 same = (!local || inet_ifa_match(local, ifa)) &&
1220 (!dst || inet_ifa_match(dst, ifa));
1224 /* Is the selected addr into dst subnet? */
1225 if (inet_ifa_match(addr, ifa))
1227 /* No, then can we use new local src? */
1228 if (ifa->ifa_scope <= scope) {
1229 addr = ifa->ifa_local;
1232 /* search for large dst subnet for addr */
1236 } endfor_ifa(in_dev);
1238 return same ? addr : 0;
1242 * Confirm that local IP address exists using wildcards:
1243 * - net: netns to check, cannot be NULL
1244 * - in_dev: only on this interface, NULL=any interface
1245 * - dst: only in the same subnet as dst, 0=any dst
1246 * - local: address, 0=autoselect the local address
1247 * - scope: maximum allowed scope value for the local address
1249 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1250 __be32 dst, __be32 local, int scope)
1253 struct net_device *dev;
1256 return confirm_addr_indev(in_dev, dst, local, scope);
1259 for_each_netdev_rcu(net, dev) {
1260 in_dev = __in_dev_get_rcu(dev);
1262 addr = confirm_addr_indev(in_dev, dst, local, scope);
1271 EXPORT_SYMBOL(inet_confirm_addr);
1277 int register_inetaddr_notifier(struct notifier_block *nb)
1279 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1281 EXPORT_SYMBOL(register_inetaddr_notifier);
1283 int unregister_inetaddr_notifier(struct notifier_block *nb)
1285 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1287 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1289 /* Rename ifa_labels for a device name change. Make some effort to preserve
1290 * existing alias numbering and to create unique labels if possible.
1292 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1294 struct in_ifaddr *ifa;
1297 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1298 char old[IFNAMSIZ], *dot;
1300 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1301 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1304 dot = strchr(old, ':');
1306 sprintf(old, ":%d", named);
1309 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1310 strcat(ifa->ifa_label, dot);
1312 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1314 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1318 static bool inetdev_valid_mtu(unsigned int mtu)
1323 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1324 struct in_device *in_dev)
1327 struct in_ifaddr *ifa;
1329 for (ifa = in_dev->ifa_list; ifa;
1330 ifa = ifa->ifa_next) {
1331 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1332 ifa->ifa_local, dev,
1333 ifa->ifa_local, NULL,
1334 dev->dev_addr, NULL);
1338 /* Called only under RTNL semaphore */
1340 static int inetdev_event(struct notifier_block *this, unsigned long event,
1343 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1344 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1349 if (event == NETDEV_REGISTER) {
1350 in_dev = inetdev_init(dev);
1352 return notifier_from_errno(-ENOMEM);
1353 if (dev->flags & IFF_LOOPBACK) {
1354 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1355 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1357 } else if (event == NETDEV_CHANGEMTU) {
1358 /* Re-enabling IP */
1359 if (inetdev_valid_mtu(dev->mtu))
1360 in_dev = inetdev_init(dev);
1366 case NETDEV_REGISTER:
1367 pr_debug("%s: bug\n", __func__);
1368 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1371 if (!inetdev_valid_mtu(dev->mtu))
1373 if (dev->flags & IFF_LOOPBACK) {
1374 struct in_ifaddr *ifa = inet_alloc_ifa();
1377 INIT_HLIST_NODE(&ifa->hash);
1379 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1380 ifa->ifa_prefixlen = 8;
1381 ifa->ifa_mask = inet_make_mask(8);
1382 in_dev_hold(in_dev);
1383 ifa->ifa_dev = in_dev;
1384 ifa->ifa_scope = RT_SCOPE_HOST;
1385 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1386 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1387 INFINITY_LIFE_TIME);
1388 inet_insert_ifa(ifa);
1393 case NETDEV_CHANGEADDR:
1394 if (!IN_DEV_ARP_NOTIFY(in_dev))
1397 case NETDEV_NOTIFY_PEERS:
1398 /* Send gratuitous ARP to notify of link change */
1399 inetdev_send_gratuitous_arp(dev, in_dev);
1404 case NETDEV_PRE_TYPE_CHANGE:
1405 ip_mc_unmap(in_dev);
1407 case NETDEV_POST_TYPE_CHANGE:
1408 ip_mc_remap(in_dev);
1410 case NETDEV_CHANGEMTU:
1411 if (inetdev_valid_mtu(dev->mtu))
1413 /* disable IP when MTU is not enough */
1414 case NETDEV_UNREGISTER:
1415 inetdev_destroy(in_dev);
1417 case NETDEV_CHANGENAME:
1418 /* Do not notify about label change, this event is
1419 * not interesting to applications using netlink.
1421 inetdev_changename(dev, in_dev);
1423 devinet_sysctl_unregister(in_dev);
1424 devinet_sysctl_register(in_dev);
1431 static struct notifier_block ip_netdev_notifier = {
1432 .notifier_call = inetdev_event,
1435 static size_t inet_nlmsg_size(void)
1437 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1438 + nla_total_size(4) /* IFA_ADDRESS */
1439 + nla_total_size(4) /* IFA_LOCAL */
1440 + nla_total_size(4) /* IFA_BROADCAST */
1441 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1442 + nla_total_size(4); /* IFA_FLAGS */
1445 static inline u32 cstamp_delta(unsigned long cstamp)
1447 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1450 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1451 unsigned long tstamp, u32 preferred, u32 valid)
1453 struct ifa_cacheinfo ci;
1455 ci.cstamp = cstamp_delta(cstamp);
1456 ci.tstamp = cstamp_delta(tstamp);
1457 ci.ifa_prefered = preferred;
1458 ci.ifa_valid = valid;
1460 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1463 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1464 u32 portid, u32 seq, int event, unsigned int flags)
1466 struct ifaddrmsg *ifm;
1467 struct nlmsghdr *nlh;
1468 u32 preferred, valid;
1470 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
1474 ifm = nlmsg_data(nlh);
1475 ifm->ifa_family = AF_INET;
1476 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1477 ifm->ifa_flags = ifa->ifa_flags;
1478 ifm->ifa_scope = ifa->ifa_scope;
1479 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1481 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1482 preferred = ifa->ifa_preferred_lft;
1483 valid = ifa->ifa_valid_lft;
1484 if (preferred != INFINITY_LIFE_TIME) {
1485 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1487 if (preferred > tval)
1491 if (valid != INFINITY_LIFE_TIME) {
1499 preferred = INFINITY_LIFE_TIME;
1500 valid = INFINITY_LIFE_TIME;
1502 if ((ifa->ifa_address &&
1503 nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1505 nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1506 (ifa->ifa_broadcast &&
1507 nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1508 (ifa->ifa_label[0] &&
1509 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1510 nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1511 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1513 goto nla_put_failure;
1515 return nlmsg_end(skb, nlh);
1518 nlmsg_cancel(skb, nlh);
1522 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1524 struct net *net = sock_net(skb->sk);
1527 int ip_idx, s_ip_idx;
1528 struct net_device *dev;
1529 struct in_device *in_dev;
1530 struct in_ifaddr *ifa;
1531 struct hlist_head *head;
1534 s_idx = idx = cb->args[1];
1535 s_ip_idx = ip_idx = cb->args[2];
1537 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1539 head = &net->dev_index_head[h];
1541 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1543 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1546 if (h > s_h || idx > s_idx)
1548 in_dev = __in_dev_get_rcu(dev);
1552 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1553 ifa = ifa->ifa_next, ip_idx++) {
1554 if (ip_idx < s_ip_idx)
1556 if (inet_fill_ifaddr(skb, ifa,
1557 NETLINK_CB(cb->skb).portid,
1559 RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1563 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1574 cb->args[2] = ip_idx;
1579 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1582 struct sk_buff *skb;
1583 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1587 net = dev_net(ifa->ifa_dev->dev);
1588 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1592 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
1594 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1595 WARN_ON(err == -EMSGSIZE);
1599 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1603 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1606 static size_t inet_get_link_af_size(const struct net_device *dev)
1608 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1613 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1616 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1618 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1625 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1629 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1630 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1635 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1636 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1639 static int inet_validate_link_af(const struct net_device *dev,
1640 const struct nlattr *nla)
1642 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1645 if (dev && !__in_dev_get_rtnl(dev))
1646 return -EAFNOSUPPORT;
1648 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1652 if (tb[IFLA_INET_CONF]) {
1653 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1654 int cfgid = nla_type(a);
1659 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1667 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1669 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1670 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1674 return -EAFNOSUPPORT;
1676 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1679 if (tb[IFLA_INET_CONF]) {
1680 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1681 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1687 static int inet_netconf_msgsize_devconf(int type)
1689 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1690 + nla_total_size(4); /* NETCONFA_IFINDEX */
1692 /* type -1 is used for ALL */
1693 if (type == -1 || type == NETCONFA_FORWARDING)
1694 size += nla_total_size(4);
1695 if (type == -1 || type == NETCONFA_RP_FILTER)
1696 size += nla_total_size(4);
1697 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1698 size += nla_total_size(4);
1699 if (type == -1 || type == NETCONFA_PROXY_ARP)
1700 size += nla_total_size(4);
1705 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1706 struct ipv4_devconf *devconf, u32 portid,
1707 u32 seq, int event, unsigned int flags,
1710 struct nlmsghdr *nlh;
1711 struct netconfmsg *ncm;
1713 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1718 ncm = nlmsg_data(nlh);
1719 ncm->ncm_family = AF_INET;
1721 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1722 goto nla_put_failure;
1724 /* type -1 is used for ALL */
1725 if ((type == -1 || type == NETCONFA_FORWARDING) &&
1726 nla_put_s32(skb, NETCONFA_FORWARDING,
1727 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1728 goto nla_put_failure;
1729 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1730 nla_put_s32(skb, NETCONFA_RP_FILTER,
1731 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1732 goto nla_put_failure;
1733 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1734 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1735 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1736 goto nla_put_failure;
1737 if ((type == -1 || type == NETCONFA_PROXY_ARP) &&
1738 nla_put_s32(skb, NETCONFA_PROXY_ARP,
1739 IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
1740 goto nla_put_failure;
1742 return nlmsg_end(skb, nlh);
1745 nlmsg_cancel(skb, nlh);
1749 void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1750 struct ipv4_devconf *devconf)
1752 struct sk_buff *skb;
1755 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1759 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1760 RTM_NEWNETCONF, 0, type);
1762 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1763 WARN_ON(err == -EMSGSIZE);
1767 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1771 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1774 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1775 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1776 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
1777 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
1778 [NETCONFA_PROXY_ARP] = { .len = sizeof(int) },
1781 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1782 struct nlmsghdr *nlh)
1784 struct net *net = sock_net(in_skb->sk);
1785 struct nlattr *tb[NETCONFA_MAX+1];
1786 struct netconfmsg *ncm;
1787 struct sk_buff *skb;
1788 struct ipv4_devconf *devconf;
1789 struct in_device *in_dev;
1790 struct net_device *dev;
1794 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1795 devconf_ipv4_policy);
1800 if (!tb[NETCONFA_IFINDEX])
1803 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1805 case NETCONFA_IFINDEX_ALL:
1806 devconf = net->ipv4.devconf_all;
1808 case NETCONFA_IFINDEX_DEFAULT:
1809 devconf = net->ipv4.devconf_dflt;
1812 dev = __dev_get_by_index(net, ifindex);
1815 in_dev = __in_dev_get_rtnl(dev);
1818 devconf = &in_dev->cnf;
1823 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1827 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1828 NETLINK_CB(in_skb).portid,
1829 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1832 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1833 WARN_ON(err == -EMSGSIZE);
1837 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1842 static int inet_netconf_dump_devconf(struct sk_buff *skb,
1843 struct netlink_callback *cb)
1845 struct net *net = sock_net(skb->sk);
1848 struct net_device *dev;
1849 struct in_device *in_dev;
1850 struct hlist_head *head;
1853 s_idx = idx = cb->args[1];
1855 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1857 head = &net->dev_index_head[h];
1859 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1861 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1864 in_dev = __in_dev_get_rcu(dev);
1868 if (inet_netconf_fill_devconf(skb, dev->ifindex,
1870 NETLINK_CB(cb->skb).portid,
1878 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1884 if (h == NETDEV_HASHENTRIES) {
1885 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
1886 net->ipv4.devconf_all,
1887 NETLINK_CB(cb->skb).portid,
1889 RTM_NEWNETCONF, NLM_F_MULTI,
1895 if (h == NETDEV_HASHENTRIES + 1) {
1896 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
1897 net->ipv4.devconf_dflt,
1898 NETLINK_CB(cb->skb).portid,
1900 RTM_NEWNETCONF, NLM_F_MULTI,
1913 #ifdef CONFIG_SYSCTL
1915 static void devinet_copy_dflt_conf(struct net *net, int i)
1917 struct net_device *dev;
1920 for_each_netdev_rcu(net, dev) {
1921 struct in_device *in_dev;
1923 in_dev = __in_dev_get_rcu(dev);
1924 if (in_dev && !test_bit(i, in_dev->cnf.state))
1925 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1930 /* called with RTNL locked */
1931 static void inet_forward_change(struct net *net)
1933 struct net_device *dev;
1934 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1936 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1937 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1938 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1939 NETCONFA_IFINDEX_ALL,
1940 net->ipv4.devconf_all);
1941 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1942 NETCONFA_IFINDEX_DEFAULT,
1943 net->ipv4.devconf_dflt);
1945 for_each_netdev(net, dev) {
1946 struct in_device *in_dev;
1948 dev_disable_lro(dev);
1950 in_dev = __in_dev_get_rcu(dev);
1952 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1953 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1954 dev->ifindex, &in_dev->cnf);
1960 static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
1962 if (cnf == net->ipv4.devconf_dflt)
1963 return NETCONFA_IFINDEX_DEFAULT;
1964 else if (cnf == net->ipv4.devconf_all)
1965 return NETCONFA_IFINDEX_ALL;
1967 struct in_device *idev
1968 = container_of(cnf, struct in_device, cnf);
1969 return idev->dev->ifindex;
1973 static int devinet_conf_proc(struct ctl_table *ctl, int write,
1974 void __user *buffer,
1975 size_t *lenp, loff_t *ppos)
1977 int old_value = *(int *)ctl->data;
1978 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1979 int new_value = *(int *)ctl->data;
1982 struct ipv4_devconf *cnf = ctl->extra1;
1983 struct net *net = ctl->extra2;
1984 int i = (int *)ctl->data - cnf->data;
1987 set_bit(i, cnf->state);
1989 if (cnf == net->ipv4.devconf_dflt)
1990 devinet_copy_dflt_conf(net, i);
1991 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1992 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
1993 if ((new_value == 0) && (old_value != 0))
1994 rt_cache_flush(net);
1996 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1997 new_value != old_value) {
1998 ifindex = devinet_conf_ifindex(net, cnf);
1999 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
2002 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2003 new_value != old_value) {
2004 ifindex = devinet_conf_ifindex(net, cnf);
2005 inet_netconf_notify_devconf(net, NETCONFA_PROXY_ARP,
2013 static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2014 void __user *buffer,
2015 size_t *lenp, loff_t *ppos)
2017 int *valp = ctl->data;
2020 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2022 if (write && *valp != val) {
2023 struct net *net = ctl->extra2;
2025 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2026 if (!rtnl_trylock()) {
2027 /* Restore the original values before restarting */
2030 return restart_syscall();
2032 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2033 inet_forward_change(net);
2035 struct ipv4_devconf *cnf = ctl->extra1;
2036 struct in_device *idev =
2037 container_of(cnf, struct in_device, cnf);
2039 dev_disable_lro(idev->dev);
2040 inet_netconf_notify_devconf(net,
2041 NETCONFA_FORWARDING,
2046 rt_cache_flush(net);
2048 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2049 NETCONFA_IFINDEX_DEFAULT,
2050 net->ipv4.devconf_dflt);
2056 static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2057 void __user *buffer,
2058 size_t *lenp, loff_t *ppos)
2060 int *valp = ctl->data;
2062 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2063 struct net *net = ctl->extra2;
2065 if (write && *valp != val)
2066 rt_cache_flush(net);
2071 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2074 .data = ipv4_devconf.data + \
2075 IPV4_DEVCONF_ ## attr - 1, \
2076 .maxlen = sizeof(int), \
2078 .proc_handler = proc, \
2079 .extra1 = &ipv4_devconf, \
2082 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2083 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2085 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2086 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2088 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2089 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2091 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2092 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2094 static struct devinet_sysctl_table {
2095 struct ctl_table_header *sysctl_header;
2096 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2097 } devinet_sysctl = {
2099 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2100 devinet_sysctl_forward),
2101 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2103 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2104 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2105 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2106 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2107 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2108 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2109 "accept_source_route"),
2110 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2111 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2112 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2113 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2114 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2115 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2116 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2117 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2118 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2119 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2120 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2121 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2122 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2123 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2124 "force_igmp_version"),
2125 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2126 "igmpv2_unsolicited_report_interval"),
2127 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2128 "igmpv3_unsolicited_report_interval"),
2130 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2131 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2132 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2133 "promote_secondaries"),
2134 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2139 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2140 struct ipv4_devconf *p)
2143 struct devinet_sysctl_table *t;
2144 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2146 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2150 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2151 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2152 t->devinet_vars[i].extra1 = p;
2153 t->devinet_vars[i].extra2 = net;
2156 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2158 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2159 if (!t->sysctl_header)
2171 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2173 struct devinet_sysctl_table *t = cnf->sysctl;
2179 unregister_net_sysctl_table(t->sysctl_header);
2183 static void devinet_sysctl_register(struct in_device *idev)
2185 neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2186 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2190 static void devinet_sysctl_unregister(struct in_device *idev)
2192 __devinet_sysctl_unregister(&idev->cnf);
2193 neigh_sysctl_unregister(idev->arp_parms);
2196 static struct ctl_table ctl_forward_entry[] = {
2198 .procname = "ip_forward",
2199 .data = &ipv4_devconf.data[
2200 IPV4_DEVCONF_FORWARDING - 1],
2201 .maxlen = sizeof(int),
2203 .proc_handler = devinet_sysctl_forward,
2204 .extra1 = &ipv4_devconf,
2205 .extra2 = &init_net,
2211 static __net_init int devinet_init_net(struct net *net)
2214 struct ipv4_devconf *all, *dflt;
2215 #ifdef CONFIG_SYSCTL
2216 struct ctl_table *tbl = ctl_forward_entry;
2217 struct ctl_table_header *forw_hdr;
2221 all = &ipv4_devconf;
2222 dflt = &ipv4_devconf_dflt;
2224 if (!net_eq(net, &init_net)) {
2225 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2229 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2231 goto err_alloc_dflt;
2233 #ifdef CONFIG_SYSCTL
2234 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2238 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2239 tbl[0].extra1 = all;
2240 tbl[0].extra2 = net;
2244 #ifdef CONFIG_SYSCTL
2245 err = __devinet_sysctl_register(net, "all", all);
2249 err = __devinet_sysctl_register(net, "default", dflt);
2254 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2255 if (forw_hdr == NULL)
2257 net->ipv4.forw_hdr = forw_hdr;
2260 net->ipv4.devconf_all = all;
2261 net->ipv4.devconf_dflt = dflt;
2264 #ifdef CONFIG_SYSCTL
2266 __devinet_sysctl_unregister(dflt);
2268 __devinet_sysctl_unregister(all);
2270 if (tbl != ctl_forward_entry)
2274 if (dflt != &ipv4_devconf_dflt)
2277 if (all != &ipv4_devconf)
2283 static __net_exit void devinet_exit_net(struct net *net)
2285 #ifdef CONFIG_SYSCTL
2286 struct ctl_table *tbl;
2288 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2289 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2290 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2291 __devinet_sysctl_unregister(net->ipv4.devconf_all);
2294 kfree(net->ipv4.devconf_dflt);
2295 kfree(net->ipv4.devconf_all);
2298 static __net_initdata struct pernet_operations devinet_ops = {
2299 .init = devinet_init_net,
2300 .exit = devinet_exit_net,
2303 static struct rtnl_af_ops inet_af_ops = {
2305 .fill_link_af = inet_fill_link_af,
2306 .get_link_af_size = inet_get_link_af_size,
2307 .validate_link_af = inet_validate_link_af,
2308 .set_link_af = inet_set_link_af,
2311 void __init devinet_init(void)
2315 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2316 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2318 register_pernet_subsys(&devinet_ops);
2320 register_gifconf(PF_INET, inet_gifconf);
2321 register_netdevice_notifier(&ip_netdev_notifier);
2323 schedule_delayed_work(&check_lifetime_work, 0);
2325 rtnl_af_register(&inet_af_ops);
2327 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2328 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2329 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
2330 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2331 inet_netconf_dump_devconf, NULL);